Uber’s brain-breaking complexity under the hood

This is a visual depicting ‘a deep dive into Uber's machine learning solution for predicting ETAs’e where ETA stands for ‘expected time of arrival’. I stumbled on this image on a Reddit thread … it struck me as deeply unsettling.

A visual deep dive into Uber's machine learning solution for predicting ETAs. Image Source

Anyone would think the expected time of arrival of an uber would be a simple thing to calculate. After all, how do you calculate your expected time of arrival when you plan to get somewhere in your city?

Yet, when you seen the visualisation of Uber’s system to accomplish this - from everyday human, non-technical eyes - you see how mind bogglingly complex these systems have already become. And to think we are only at the dawn a new age of generative and spatial intelligence combined with the next many orders of magnitude in leveraging big data.

Essentially, it seems the more such systems become ontologically complex, the more their vaneer is of simplicity, ease, convenience. Indeed, one begets the other. The hybrid of these inner and outer dimensions when combined is experienced more as a living feedback loop, an ever evolving prosthetic that is increasingly more responsive to the whims, fears, delights and behaviors of Uber’s users.

Makes me think of John Maeda’s quote, “Simplicity and complexity need each other.” In this case, the tacit alliance is between the customer and the product, where the customer chooses to increasingly distance and deny to themselves the extent of the mind-boggling complexity it takes for them to get the convenience they do from the product. In exchange for this perceived simplicity, they also hand over participating in the price negotiation, instead freely handing over their inputs into the system in terms of their behavioral data.

I’m thinking that it would be fun to grab a random person and ask them to sketch out a visual of how they think / model how Uber’s ETA prediction system works. Watch this space …